Machine Learning Engineer (F/M) – Internship

Why this role is key

The digital twins of our antennas are used to configure our technology to steer electromagnetic waves. These models combine a physical understanding of our antennas with trained parameters. This internship aims to further improve the robustness and reliability of our model training.

Reporting to Steph, Software Engineer, you will investigate the factors that can influence training stability and evaluate different approaches to improving it. Working with our Algorithms and Modelling, Radio Frequency and Software teams, you will learn to connect physical modelling, numerical optimisation and experimental analysis through controlled, reproducible comparisons.

Your responsibilities

Understanding the model and its training process

• Understand our physics-based antenna models and how they are trained,

• Reproduce a training run and measure the variability in the resulting performance.

Diagnosing and improving stability

• Set up monitoring of relevant quantities and automatically detect diverging training runs,

• Investigate and implement projection, reparameterisation or penalty methods to enforce model convergence at each optimisation step,

• Evaluate the impact of these methods on accuracy and computation time.

Optimising and comparing training runs

• Improve stopping criteria, initialisation and the reuse of previously trained models,

• Compare several model variants under controlled conditions,

• Analyse the model’s physical and numerical mechanisms to investigate their influence on the results,

Documentation and reproducibility

• Test the code developed and maintain it under version control,

• Document the data and training conditions to ensure reproducibility of the results,

• Present the methods investigated and the results obtained in reports supported by quantitative data.